METHOD FOR CONTROLLING AND POSSIBLY RECOVERING THE ADHESION OF THE WHEELS OF A CONTROLLED AXLE OF A RAILWAY VEHICLE

Abstract

A method for controlling and recovering the adhesion, during a slipping phase, of wheels (W.sub.i) belonging to at least two controlled axles (A.sub.i) of a railway vehicle, comprising the steps of: generating speed signals indicative of the angular speed (?.sub.i) of said wheels (W.sub.i); estimating the value of the instantaneous adhesion (?(T.sub.j)) at the point of contact of such wheels (W.sub.i) and the rails, using an adhesion observer; generating a target-slip value (?) for the wheels (W.sub.i) of the controlled axles (A.sub.i) by means of an optimization algorithm which processes the estimated adhesion values (?.sub.i(T.sub.j)), and modifying the target-slip value continuously in time, with a predetermined sampling period, such as to maximize the average value of the adhesion of the wheels of the vehicle.

Claims

1. A method for controlling and recovering the adhesion of wheels (W.sub.i) belonging to at least two controlled axles (A.sub.i) of a railway vehicle, during a wheel skidding phase, comprising the steps of: generating speed signals indicative of the angular speed (?.sub.i) of said wheels (W.sub.i); estimating the value of the instantaneous adhesion (?(T.sub.j)) at the point of contact of said wheels (W.sub.i) and the rails, using an adhesion observer; generating a target-slip value (?) for the wheels (W.sub.i) of said at least two axles (A.sub.i) by means of an optimization algorithm which processes the estimated adhesion values (?.sub.i(T.sub.j)), and modifying said target-slip value continuously in time, with a predetermined sampling period (T), such as to maximize the average value of the adhesion of the wheels of the vehicle.

2. A method according to claim 1, wherein the derivative $\left(\frac{d.Math..Math.\stackrel{\_}{?}}{d.Math..Math.?}\right)$ of the average value (?) of the adhesion of the wheels (W.sub.i) of said axles (A.sub.i) as a function of the variation of the target-slip (?) of said wheels (W.sub.i) is computed, and through an adaptive filter the value of the target-slip (?) for a subsequent sampling interval (T.sub.j+1) is modified, such as to make said derivative $\left(\frac{d.Math..Math.\stackrel{\_}{?}}{d.Math..Math.?}\right)$ go to zero, i.e. such as to maximize the average value of the adhesion (?) of the wheels (W.sub.i) of said axles (A.sub.i).

3. A method according to claim 1, wherein said derivative $\left(\frac{d.Math..Math.\stackrel{\_}{?}}{d.Math..Math.?}\right)$ is computed and thereafter integrated by means of an integrator the output of which modifies the value of the target-slip (?) and becomes stable when said derivative is equal to zero, i.e. when the average value of the adhesion (?) of said axles (A.sub.i) tends to its maximum value.

4. A method according to claim 1, wherein the sign of said derivative $\left(\frac{d.Math..Math.\stackrel{\_}{?}}{d.Math..Math.?}\right)$ is computed and thereafter integrated by means of an integrator the output of which modifies the value of the target-slip (?) and becomes stable when said derivative is equal to zero, i.e. when the average value of the adhesion (?) of said axles (A.sub.i) reaches its maximum value.

5. A method according to claim 1, wherein the value of the difference between the maximum adhesion (?.sub.max(T.sub.j)) and the minimum adhesion (?.sub.min(T.sub.j)) of the wheels (W.sub.i) of said controlled axles (A.sub.i) is computed and through a predetermined function with hysteresis, which links the target-slip (?) with the adhesion (?), the value of the target-slip (?) is modified and stabilized about the maximum value of the average adhesion (?).

6. A method according to claim 1, wherein the average value of the adhesion (?), the maximum variation of the adhesion (?.sub.max-?.sub.min) and the target-slip (?(T.sub.j)) are computed, and through fuzzy logic devices a subsequent value of target-slip, ?(T.sub.j+1) is generated, to be assigned to an adhesion recovery module.

7. A method according to claim 1, wherein the last skidding axle, with respect to the direction of travel, is maintained in a condition of controlled slipping at the peak value of the available adhesion.

8. A method according to claim 1, wherein the vehicle speed (VV) is computed by means of the instantaneous speed of at least one axle which is skidding, said axle being kept in a condition of controlled slipping at the peak value of the available adhesion.

9. A method according to claim 1, wherein at least two axles which are skidding are maintained in a condition of controlled slipping at the peak value of the available adhesion for determining the direction of travel of the vehicle.

10. A method according to claim 7, wherein in order to maintain an axle in a condition of skidding at a peak value of the available adhesion, a control algorithm for said axle is used, wherein on the basis of the adhesion value (?) for that axle, the derivative $\left(\frac{d.Math..Math.?}{d.Math..Math.?}\right)$ of the adhesion (?) as a function of the slip (?) is computed, and through an adaptive filter the torque value (C) to be assigned to a system for controlling the torque applied to the axle is modified, such as to keep said derivative substantially at zero.

11. A method according to claim 7, wherein in order to keep said axles in a condition of skidding at the peak value of the adhesion, a control algorithm for each axle is used, wherein the sign of the derivative $\left(\frac{d.Math..Math.?}{d.Math..Math.?}\right)$ of the adhesion (?) as a function of the slip (?) is computed and thereafter integrated by means of an integrator, the output of which modifies the torque value (C) to be assigned to a system for controlling the torque applied to the axle, such as to make said derivative substantially equal to zero.

12. A method according to claim 1, applied during a condition of skidding in a traction phase, or in a condition of slipping in a braking phase.

Description

[0040] Further features and advantages of the invention will become apparent from the detailed description that follows, provided by way of non-limiting example with reference to the accompanying drawings, in which:

[0041] FIG. 1 is a block diagram of an anti-skid control system of the wheels of a railway vehicle;

[0042] FIG. 2 is a block diagram of a closed loop control system of an axle's rotation speed;

[0043] FIG. 3 is a diagram of a possible embodiment of an apparatus for controlling the torque applied to an axle;

[0044] FIG. 4 is a graph showing qualitatively the trend of the adhesion coefficient t of the wheels of an axle, shown on the y-axis, as a function of the slip ?, shown on the x-axis;

[0045] FIG. 5 is a diagram illustrating the forces applied to an axle's wheel;

[0046] FIGS. 6A, 6B are graphs showing qualitatively the trends of the adhesion coefficient ? of the wheels of four axles of a vehicle in two different operating conditions;

[0047] FIG. 6C illustrates the trend of an average adhesion curve ? around the peak value;

[0048] FIG. 7 is a block diagram of a system for implementing a method according to the present invention;

[0049] FIGS. 8 and 9 are block diagrams of two alternative embodiments of systems for continuously tracking the average adhesion peak value;

[0050] FIG. 10 is a block diagram of another system for implementing a process or method according to the present invention;

[0051] FIG. 11 is a graph of a transfer function with hysteresis used in the implementation of the method according to the invention; and

[0052] FIG. 12 is a block diagram of a variant of embodiment of a system for implementing the method according to the present invention;

[0053] As will appear more clearly from the following, the method according to the present invention allows the optimum value of the slip ?(t) to be identified, which allows the adhesion value obtained as the average value between the instantaneous adhesion of all the axles to be maximized, this average value being defined as follows:

?(t)=?.sub.1.sup.n?.sub.n(?,t)/n i=1,2, . . . ,n(7)

[0054] The method according to the present invention intervenes at the beginning of a skidding phase and corrects said optimum value of ?(t) in real time and continuously over time, adapting it to the possible variations of the values ?.sub.i(?,t) (adhesions of the i controlled axles) which may intervene in the course of skidding so as to tend to maintain the average value ?(t) in all circumstances at the maximum value.

[0055] The method according to the present invention uses an adhesion observer to evaluate in real time the adhesion value ? at the point of contact between the wheels and rails for one or more axles during a skidding phase and, by processing these ? values in real time, identifies continuously over time the optimal ? value to be assigned to a slip control system to obtain the greatest global adhesion recovery.

[0056] An adhesion observer adapted to dynamically identify the instantaneous value (T.sub.j) of the adhesion in a generic sampling period T.sub.j of a predetermined duration T at the wheel-rail point of contact during skidding is definable using the equations provided above, from which with some simple steps the following relationship is obtained:

[00002]$\begin{array}{cc}??\left(T_j\right)=\frac{1}{m.Math.g}.Math.\left[F_m?\left(T_j\right)+J/R.Math.\stackrel{.}{?}?\left(T_j\right)\right]& \left(8\right)\end{array}$

where [0057] {dot over (?)} is the angular acceleration of the axle, i.e. the time derivative of the angular speed co of the axle; the value of this acceleration is already available in real time within a control and adhesion recovery system, because angular acceleration is one of the variables on which the control function implemented by the block CM of FIG. 2 is normally based for achieving the slip control of the axle; the sign of {dot over (?)} depends on the instantaneous acceleration or deceleration condition of the axle; [0058] m is the mass on the wheel-rail point of contact; in the latest generation trains, the m value is known in real time, as it is commonly available to the system that computes the accelerating/braking force to apply to the axle to obtain the desired accelerations/decelerations; [0059] J is the moment of inertia of the axle and is a parameter whose value is always known, being supplied by the manufacturer of the carriages, as it is fundamental for the computation of stopping distances; [0060] F.sub.m, already defined above in relation to FIG. 5, can be obtained by multiplying the pressure applied to the brake cylinder, known to the braking system, for pressure/force conversion coefficients typical of the brake cylinder, as well as the transmission and efficiency coefficients of the levers and of the coefficient of the friction between the brake linings and discs (in the case of disc brakes); in the case of electrodynamic type traction or braking, the value of the force F.sub.m may be obtained from the electric current value supplied/regenerated by the motor in traction or, respectively, in braking; in the case of so-called blended braking, the intensity of the force F.sub.m may be determined as the sum of the respective contributions of the pneumatic brake and of the electrodynamic brake, appropriately weighed with respective coefficients; and [0061] T.sub.j is the generic j-th value for the sampling period of the system with which the adhesion observer and more generally the method according to the invention is carried out; in the description that follows, T.sub.j will replace the use of the variable t representing time.

[0062] Downstream of the adhesion observer, a low-pass type filter may appropriately be provided, to remove or at least mitigate instantaneous variations and noise present outside of the frequency band useful for a correct observation of the adhesion values.

[0063] A first embodiment of a system for implementing a method according to the present invention is illustrated in FIG. 7.

[0064] The method provides for identifying and tracking the slip value ? such that the curve ?(?) illustrated in FIG. 6C presents the maximum value, i.e. the ? value for which

[00003]$\frac{d.Math..Math.\stackrel{\_}{?}.Math..Math.\left(T\right)}{d.Math..Math.??\left(T\right)}=0.$

[0065] For this purpose, a system implementing an LMS algorithm (Least Mean Square) may be used. For an accurate description of the general characteristics of the convergence criteria and the implementation variants of LMS algorithms, please refer to the available literature and in particular to the text: B. Widrow, S. D. Stearns, Adaptive Signal Processing, New Jersey, Prentice-Hall, Inc., 1985.

[0066] With reference to FIG. 7, an adhesion observer 701 receives input signals representative of the speed values of ?1, ?2, . . . , ?n of the wheels W1, W2, . . . , Wn of controlled axles A1, A2, . . . , An, together with a vector containing the values of the magnitudes m.sub.i(T.sub.j), J.sub.i, R.sub.i and F.sub.mi(T.sub.j) previously described for the estimation of the instantaneous adhesion values of ?.sub.i(T.sub.j) relating to the axles A.sub.i (with i=1, 2, . . . , n).

[0067] The output of the adhesion observer 701 is connected to the input of a module 702 which computes, based on the estimated instantaneous adhesions values ?.sub.i(T.sub.j), the average value ?(T.sub.j).

[0068] A subsequent differentiator module 703 computes the value of

[00004]$\frac{d.Math..Math.\stackrel{\_}{?}}{d.Math..Math.?},$

for example, according to the equation:

[00005]$\begin{array}{cc}\frac{d.Math..Math.\stackrel{\_}{?}.Math..Math.\left(T_j\right)}{d.Math..Math.??\left(T_j\right)}=\frac{\stackrel{\_}{?}?\left(T_j\right)-\stackrel{\_}{?}?\left(T_{j-1}\right)}{??\left(T_j\right)-??\left(T_{j-1}\right)}& \left(9\right)\end{array}$

[0069] An adder 704 outputs the error e(T.sub.j) as the difference between the desired value (0) of said derivative and its instantaneous value corresponding to the equation (9) given above. The error e(T.sub.j) is used to drive and adapt the LMS algorithm implemented in a block 705. This block outputs the target value ?(T.sub.j+1).

[0070] The value ?(T.sub.j+1) is supplied, together with the updated value of the speed Vv of the vehicle, to a plurality of adhesion recovery control blocks 706, one for each axle A.sub.i, each having, for example, the architecture illustrated in FIG. 2 described above.

[0071] The module 705 that implements the LMS algorithm continuously implements the correction of the output, i.e. the ? value, in order to minimize or cancel the error e(T), i.e. up to the cancellation of

[00006]$\frac{d.Math..Math.\stackrel{\_}{?}}{d.Math..Math.?}.$

A simplified implementation of the group of modules included in the dashed line block 710 of FIG. 7 is illustrated in FIG. 8, where the block 705, which implements the LMS algorithm, is replaced with a simple integrator 805, the output of which, amplified with a gain K, generates the target-slip value ?(T.sub.j+1) to be assigned to the adhesion control and recovery system 706 (FIG. 7).

[0072] The gain K regulates the identification speed of the average adhesion peak value f and simultaneously ensures the stability of the closed loop system.

[0073] A further simplified variant of embodiment of the dashed block 710 of FIG. 7 is shown in FIG. 9: the module 903 detects the sign of the derivative

[00007]$\frac{d.Math..Math.\stackrel{\_}{?}}{d.Math..Math.?}.$

The output of the block 903 being equal to +1 or ?1 (the positive and, respectively, negative direction), a subsequent integrator 805 performs simple unitary sums. The integrator 805 may be replaced with an up/down type counter updated with period T=T.sub.j+1?T.sub.j.

[0074] The diagrams according to FIGS. 8 and 9 perform a continuous tracking of the average adhesion peak ?, continuously adapting to the change in adhesion conditions, similarly to what was achieved with the diagram according to FIG. 7. The latter allows rapid and accurate tracking of the condition

[00008]$\frac{d.Math..Math.\stackrel{\_}{?}}{d.Math..Math.?}=0,$

but requires the use of a certain number of computations in real time.

[0075] The diagram according to FIG. 9 greatly reduces the number of computations necessary, but also reduces the speed of tracking the condition

[00009]$\frac{d.Math..Math.\stackrel{\_}{?}}{d.Math..Math.?}=0.$

[0076] The diagram according to FIG. 8 has intermediate characteristics between those of the diagrams according to FIGS. 7 and 9.

[0077] FIG. 10 illustrates a further system for the implementation of a method according to the present invention, wherein the difference between the greater and lesser adhesion value between the controlled axles in the generic period T.sub.j is analyzed in real time:

??(T.sub.j)=?.sub.max(T.sub.j)??.sub.min(T.sub.j)(10)

and the value ?(T.sub.j+1) is obtained on the basis of a curve obtained from experimental data, as better described below.

[0078] With reference to FIG. 10, an adhesion observer 1001, similar to the observer 701 of FIG. 7, receives the values of the speeds ?.sub.i of the wheels W.sub.i of controlled axles A.sub.i, together with a vector of the values of the magnitudes previously described, necessary for the estimation of the corresponding adhesions ?.sub.i(T.sub.j). A module 1002 receives from the adhesion observer 1001 the values of the instantaneous adhesions ?.sub.i(T.sub.j) and outputs the value of ??(T.sub.j), according to the equation (10) given above.

[0079] A subsequent module 1003 receives as input the value of ??(T.sub.j) and outputs the value of ?(T.sub.j+1) to be assigned to the control and adhesion recovery module 1004, similar to the module 706 of FIG. 7 and having, for example, the configuration shown in FIG. 2.

[0080] Appropriately, the module 1003 may have a transfer function with hysteresis according to the graph shown in FIG. 11: this transfer function defines a relationship between the slip ? and the adhesion variation ??, the graph of which has essentially the shape of a polygon, bounded below by a straight horizontal line, ?=?.sub.x with ?.sub.x typically (but not necessarily) equal to 0.05 and bounded above by a horizontal straight line, ?=?.sub.y, with ?.sub.y typically (but not necessarily) equal to 0.35. The transfer function can thus generate ? values between ?.sub.x and ?.sub.y.

[0081] If the adhesion control and recovery module 1004 must fully comply with regulatory requirements (EN 15595, :2009+A1, cited above), then the ?.sub.y value must abide by the requirements in paragraph 6.3.2.2 of said standard.

[0082] If during a sliding phase for a given ? value, a reduction of adhesion ?? is observed tending to cause the point of work to migrate out horizontally through the left oblique rectilinear side of the aforementioned polygon, the transfer function will determine the new value of ?(??) descending along this oblique rectilinear side. Similarly, if, during a skidding phase for a given ? value, there is an increase of ?? tending to cause the point of work to migrate out horizontally through the right oblique side of the polygon, the transfer function will determine the new value of ?(??) rising along the right oblique rectilinear side of the aforementioned polygon.

[0083] The hysteresis of the transfer function is required to provide stability to the system, which otherwise would tend to oscillate due to the significant propagation delay in the loop.

[0084] The oblique rectilinear sides of the polygon converge between them toward the bottom, reducing the hysteresis in the vicinity of the origin of the coordinate axes, in order to make the system very sensitive to small variations of ?? when the system is to work in conditions of ???.sub.x, as in the situation to which the graph of FIG. 6A refers.

[0085] In FIG. 11, the values p, q, r, which represent the x-coordinates of vertices of the aforementioned polygon, are determined experimentally and have for example approximately the values p=0.01, q=0.03 and r=0.05.

[0086] The module 1003 computes ?(T.sub.j+1) with a period T (=T.sub.j+1?T.sub.j), ensuring an adjustment in time of the ? value to the environmental conditions.

[0087] A further implementation of the method according to the present invention may provide for the generation of the value of ?(T.sub.j) according to a real-time processing of the values of ?(T.sub.j), ??(T.sub.j) and ?(T.sub.j) by means of a fuzzy logic algorithm, intended to generate the value of ?(T.sub.j+1) to be assigned to the adhesion control/recovery module 706 of FIG. 7.

[0088] Each manner of implementing the method according to the invention described above in skidding phase forces all the controlled axles to slip about the value ?. In fact, the last (in the direction of travel) of the controlled axles that is still in the skidding condition, no longer having the function of cleaning the rails for any subsequent axles (since it is the last of the axles, or further subsequent axles being in the condition of complete adhesion) may be held in controlled slipping on the adhesion peak value lying on the curve A of FIG. 4 by further increasing the value of ?(T.sub.j).

[0089] Such action simply cannot be done by forcing on the concerned axle a specific value of ? corresponding to the points of the curve A of FIG. 4, since this curve is unknown a priori and varies continuously with time.

[0090] To maintain this axle in controlled slipping on the adhesion peak value, as is shown in FIG. 12, an adhesion observer 1201 receives signals indicative of the wheel speed W of this controlled axle, simultaneously with a value vector of the magnitudes, previously described, necessary for estimating the instantaneous adhesion ? of this axle.

[0091] A subsequent module 1202 computes the value of the derivative

[00010]$\frac{d.Math..Math.?}{d.Math..Math.?},$

when the value of ? is obtained in real time in accordance with the equation (1).

[0092] An adder 1203 outputs the error e(T.sub.j) as the difference between the desired value of said derivative (i.e., the value 0) and the instantaneous value computed by the module 1202. This error is used to adapt the LMS algorithm implemented in a block 1204. The latter outputs a torque request C(T.sub.j+1) for said axle, which is transmitted to a torque control module 1205, having, for example, the architecture described above with reference to FIG. 3.

[0093] In a manner known per se, the module 1204 continuously corrects the output C(T.sub.j+1) in order to minimize or cancel the error e(T), i.e. in order to obtain a cancellation of the aforementioned derivative, that is in order to bring said axle to the adhesion peak value and maintain it there.

[0094] The dashed block 1206 of FIG. 12 may possibly be simplified as described above in relation to FIG. 7 and the relative simplifying variants illustrated in FIGS. 8 and 9.

[0095] The solution according to FIG. 12 allows the real value of the maximum available adhesion for a given axle to be measured.

[0096] By applying this solution to two axles, for example, the first axle in the direction of travel and the last axle in the skidding condition, and finding the difference between their adhesions, the value to be assigned as the difference in adhesion ?? in the embodiment illustrated in FIG. 10 may be obtained, in substitution of the blocks 1001 and 1002 illustrated herein.

[0097] The solution according to FIG. 12 may also be used to identify the direction of travel of the vehicle: at the beginning of a skidding phase, the solution according to FIG. 12 is applied for example to the first and last axles of the vehicle and the direction of travel is defined by the axle for which the lower value of adhesion is detected.

[0098] Finally, the solution according to FIG. 12 may be used to improve the estimation of the actual speed Vv of the vehicle: in fact, the curve A of FIG. 4 is located in a field to which correspond the x-axis values ? of less than 0.02. The algorithm most used for the estimation of the actual speed Vv of the vehicle, in the event of braking, normally uses a function of the type:

V.sub.v(T.sub.j)=max[S.sub.1(T.sub.j), . . . ,S.sub.n(T.sub.j),(V.sub.v(T.sub.j?1)+a.sub.max.Math.T)](11)

while in case of traction, the following function is used:

V.sub.v(T.sub.j)=min[S.sub.1(T.sub.j), . . . ,S.sub.n(T.sub.j),(V.sub.v(T.sub.j?1)+a.sub.max.Math.T)](12)

where a.sub.max is the maximum acceleration permitted for the vehicle in operation, this acceleration having a positive sign in the case of a traction condition and a negative sign in the case of a braking condition.

[0099] Therefore, applying the solution according to FIG. 12 to at least one axle, said axle will always advance at a linear speed equal to that of the vehicle (less than a maximum error computable as 2%), even in degraded adhesion conditions. Therefore, the above two expressions provided always allow a very reliable value of the vehicle's speed Vv to be provided, even in very degraded adhesion conditions.

[0100] Naturally, without altering the principle of the invention, the embodiments and the details of implementation may vary widely with respect to those described and illustrated purely by way of non-limiting example, without thereby departing from the scope of the invention as defined in the appended claims.